general_cfg:
  algo_name: PPO
  collect_traj: false
  device: cuda
  env_name: gym
  load_checkpoint: true
  load_model_step: best
  load_path: Train_single_Hopper-v4_PPO_20230526-135633
  max_episode: 100
  max_step: 1000
  mode: test
  model_save_fre: 500
  mp_backend: single
  n_learners: 1
  n_workers: 2
  online_eval: true
  online_eval_episode: 10
  seed: 1
  share_buffer: true
algo_cfg:
  actor_hidden_dim: 256
  actor_layers:
  - activation: relu
    layer_size:
    - 256
    layer_type: linear
  - activation: relu
    layer_size:
    - 256
    layer_type: linear
  actor_lr: 0.0003
  batch_size: 256
  buffer_size: 100000
  buffer_type: ONPOLICY_QUE
  continuous: true
  critic_hidden_dim: 256
  critic_layers:
  - activation: relu
    layer_size:
    - 256
    layer_type: linear
  - activation: relu
    layer_size:
    - 256
    layer_type: linear
  critic_loss_coef: 0.5
  critic_lr: 0.001
  entropy_coef: 0.01
  eps_clip: 0.2
  epsilon_decay: 500
  epsilon_end: 0.01
  epsilon_start: 0.95
  gamma: 0.95
  independ_actor: true
  k_epochs: 8
  kl_alpha: 2
  kl_beta: 1.5
  kl_lambda: 0.5
  kl_target: 0.1
  lr: 0.0001
  min_policy: 0
  ppo_type: clip
  sgd_batch_size: 128
  share_optimizer: false
  target_update: 4
env_cfg:
  id: Hopper-v4
  ignore_params:
  - wrapper
  - ignore_params
  new_step_api: true
  render_mode: null
  wrapper: null